Several years ago, a large retailer tried to encourage word-of-mouth marketing for products sold on its site by offering incentives to site visitors who made product recommendations. Many companies are trying to use people’s social connections for such “viral marketing” programs, hoping that information about products (and the urge to buy them) can spread through a network of people the way a virus might. But after studying more than 15 million recommendations generated by the retailer’s incentive program, a team made up of Jure Leskovec, Lada Adamic, and Bernardo Huberman, director of the information dynamics lab at Hewlett-Packard, was skeptical. Huberman and his colleagues looked at the networks that grew up around each product–who bought and recommended it, and who responded to the recommendation–and saw that they took on different characteristics depending on the type of product. A network around a medical book (top image below), where red dots and lines indicate people who purchased the book while blue dots and lines represent people who received a recommendation, shows a scattered network where recommendations, on average, don’t travel very far. The network surrounding a Japanese graphic novel (bottom image below), on the other hand, shows a thick flow of information among densely connected people. The researchers found that viral marketing was most effective for expensive products recommended within a small, tightly connected group. They also found that overusing consumers’ social connections for marketing can make them less influential.
Credit: Jure leskovec, Lada Adamic, Bernardo Huberman, from “The Dynamics of Viral Marketing,” ACM Transactions on the Web, Vol. 1, Issue 1, May 2007